Diverse Diazotrophs Are Present on Sinking Particles in the North Pacific

The ISME Journal https://doi.org/10.1038/s41396-018-0259-x

ARTICLE

Diverse diazotrophs are present on sinking particles in the North Paciﬁc Subtropical Gyre

1,2 1 3 4 4,5,6 Hanna Farnelid ● Kendra Turk-Kubo ● Helle Ploug ● Justin E. Ossolinski ● James R. Collins ● 4 1 Benjamin A. S. Van Mooy ● Jonathan P. Zehr

Received: 29 March 2018 / Revised: 3 July 2018 / Accepted: 26 July 2018 © The Author(s) 2018. This article is published with open access

Abstract Sinking particles transport carbon and nutrients from the surface ocean into the deep sea and are considered hot spots for bacterial diversity and activity. In the oligotrophic oceans, nitrogen (N2)-ﬁxing organisms (diazotrophs) are an important source of new N but the extent to which these organisms are present and exported on sinking particles is not well known. Sinking particles were collected every 6 h over a 2-day period using net traps deployed at 150 m in the North Paciﬁc Subtropical Gyre. The bacterial community and composition of diazotrophs associated with individual and bulk sinking particles was assessed using 16S rRNA and nifH gene amplicon sequencing. The bacterial community composition in bulk

1234567890();,: 1234567890();,: particles remained remarkably consistent throughout time and space while large variations of individually picked particles were observed. This difference suggests that unique biogeochemical conditions within individual particles may offer distinct ecological niches for specialized bacterial taxa. Compared to surrounding seawater, particle samples were enriched in different size classes of globally significant N2-fixing cyanobacteria including Trichodesmium, symbionts of diatoms, and the unicellular cyanobacteria Crocosphaera and UCYN-A. The particles also contained nifH gene sequences of diverse non- cyanobacterial diazotrophs suggesting that particles could be loci for N2 fixation by heterotrophic bacteria. The results demonstrate that diverse diazotrophs were present on particles and that new N may thereby be directly exported from surface waters on sinking particles.

Introduction

Marine environments exhibit microscale gradients of organic carbon and nutrients, representing diverse micro- Electronic supplementary material The online version of this article niches for microorganisms to exploit. Particulate organic (https://doi.org/10.1038/s41396-018-0259-x) contains supplementary material, which is available to authorized users. matter (POM) is ubiquitous in seawater and composed of a mixture of living microorganisms and dead detrital matter, * Hanna Farnelid providing ‘hotspots’ for microbial activity and organic [email protected] matter remineralization that signiﬁcantly contribute to bio- 1 Ocean Sciences Department, University of California at Santa geochemical element cycling rates [1]. Sinking particles Cruz, Santa Cruz, CA, USA transport organic and inorganic material into the deep ’ 2 Centre for Ecology and Evolution in Microbial Model Systems, ocean, a process known as the ocean s biological pump [2]. Linnaeus University, Kalmar, Sweden Because of their importance, sinking particles and the 3 Department of Marine Sciences, University of Gothenburg, composition and function of their associated community is Gothenburg, Sweden receiving increasing interest [3–5]. Compared to larger 4 Department of Marine Chemistry & Geochemistry, Woods Hole phytoplankton, the representation of picoplankton in sink- Oceanographic Institution, Woods Hole, MA, USA ing particles has generally been thought to be negligible due 5 MIT/WHOI Joint Program in Oceanography, University of to their small size, slow sinking velocity and tight regulation Washington, Seattle, WA, USA through grazing [6]. However, in recent years the possibility 6 School of Oceanography and eScience Institute, University of of export of small picoplankton through incorporation into Washington, Seattle, WA, USA aggregates or fecal pellets has been increasingly recognized H. Farnelid et al.

[7, 8] and has been proposed to be in proportion to their while UCYN-B (Crocosphaera sp.) can be found free- total net primary production [9, 10]. living but may also form colonies, aggregates or live in The bacterial community composition of particulate symbiosis with a diatom [26]. The physiology of the material has been assessed in several different environ- broadly defined UCYN-C group is still largely unknown. ments, suggesting that particle attached communities are To explore diazotroph diversity, the nifH gene, a gene phylogenetically distinct from that of the surrounding water encoding a subunit of the nitrogenase enzyme, has been (e.g., [5, 11–13]). Groups that are enriched on sinking frequently used. In recent years nifH gene sequences related particles include members of Bacteroidetes, Planctomy- to non-cyanobacterial, heterotrophic diazotrophs have been cetes, and Roseobacter [3, 14]. Since particles are thought reported from various oligotrophic ocean regions [27, 28]. to represent diverse microscale conditions, bulk analyses of How heterotrophic diazotrophs support N2 fixation in the sedimented or size fractionated material may not provide oxygenated surface ocean is not yet understood and the high-resolution insights into the associated communities of significance to global N2 fixation rates is debated [29, 30]. It individual particles. In addition, only few studies have has been proposed that heterotrophic N2 fixation may take considered degradation dynamics during sample collection place in close association with phytoplankton or within [15]. Thus, molecular characterization of individual parti- interiors of particles but this has not yet been resolved [29]. cles collected during short time scales may provide novel Sinking particles constitute a diverse spectrum of envir- insights into the composition of sinking particles. onments for their colonizers based on the history of the The North Pacific Subtropical Gyre (NPSG) is char- organic matter by which they are formed. Low O2 con- acterized by low concentrations of macronutrients and centrations within sinking particles and aggregates due to warm surface water temperatures. The ecosystem is domi- microbial remineralization of organic matter have pre- nated by relatively low rates of primary production and low viously been reported [31, 32]; these microenvironments particulate organic carbon (POC) flux at the base of the would be potential niches for O2-sensitive diazotrophs. In euphotic zone [16]. Primary production in the NPSG is this study, we investigate the bacterial composition and the limited by nitrogen (N) and the dissolved dinitrogen gas associated populations of diazotrophs on both individually (N2) is only available to N2-fixing organisms (diazotrophs), picked and bulk samples of sinking particles in the NPSG, which can reduce N2 into bioavailable ammonium. In the and compare the community composition with that of the NPSG, diazotrophs contribute to a significant portion surrounding seawater. Using high-throughput nifH gene (26–47% during the years 2006–2013) of particulate nitro- amplicon sequencing we also investigate whether diazo- gen export [17]. However, the fate of the newly fixed N2 in troph biomass is exported on sinking particles and if par- the marine food web is not fully understood and the extent ticles could be potential loci for heterotrophic N2 fixation in to which cells of dead or living N2-fixing organisms are the oligotrophic ocean. exported from the euphotic zone on sinking particles may vary among the diverse species [18, 19]. In the oligotrophic open ocean, significant diazotrophs associated with larger Methods size fractions of plankton include the widely distributed filamentous N2-fixing cyanobacterium Trichodesmium [20] Sample collection at sea and heterocystous-cyanobacterial endosymbionts of diatoms [21]. While diatoms can be exported directly due to Lagrangian sampling was performed during the HOE- their relatively heavy silicified cell walls [22] Tricho- Legacy 2B cruise (KOK1507) in the NPSG onboard the R/ desmium has been suggested to have low direct export VKa’imikai-O-Kanaloa in an anticyclonic eddy (Fig. 1). efficiency [23]. Small unicellular cyanobacterial diazo- Water column properties (0–200 m) were determined from trophs (UCYN), are also highly prevalent (e.g., [24]) and vertical depth profiles using a rosette of Niskin bottles may consequently play a significant role in POC export, but equipped with a fluorometer and temperature, conductivity little is known about their possible export mechanisms. and oxygen sensors every 4 h during 25 July to 3 August N2 fixation is an energy demanding and oxygen (O2) 2015. Surface-tethered particle net traps with a diameter of sensitive process. Cyanobacterial diazotrophs have various 1.25 m (1.23 m2 area of collection) and a 20 µm mesh cod mechanisms for protecting the nitrogenase enzyme from end [33] were deployed at 150 m for ~5–6 h. A total of 7 inactivation by O2 including temporal separation of N2 deployments (Fig. 1,D1–D7) were conducted over a 48 h fixation and photosynthesis, heterocyst formation, and period (Table S1). An acoustic release was used to close the symbiotic interactions. The unicellular cyanobacterial dia- traps prior to recovery so that material from the overlying zotrophs (UCYN-A, UCYN-B and UCYN-C) exhibit a water column was not inadvertently sampled. Traps were range of cellular strategies and interactions [25]. UCYN-A recovered immediately upon closure, except in the case of lives in association with a photosynthetic picoeukaryote, net trap D4, which was closed after ~6 h but could not be Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre

Fig. 1 Map showing the sampling location of the net traps (D1–D7) which were collected within a total distance of 16 km, 150 m seawater samples (S1 and S2) and vertical depth proﬁle collected for nifH community analysis located in an anticyclonic eddy north of Hawaii adapted from Wilson et al. [50]. a Patterns of sea level anomaly. b Sampling locations

recovered until ~13 h after it was deployed (Table S1). (Sterlitech, Kent, WA, USA) and subsequently though a Particle material was divided in acid-rinsed 500 ml HDPE 0.2 μM Supor filter in 25 mm Swinnex filter holders (Mil- bottles as described in Collins et al. [34] and processed lipore, Billerca, MA, USA) using a peristaltic pump. Sam- immediately. Bulk particle samples were obtained by fil- ples (2 l) were also collected from a vertical seawater profile tering 10 ml of the material collected in the traps through a (5, 15, 25, 35, 45, 60, 75, and 100 m; Fig. 1) on 0.2 μM 0.2 µm pore size Supor filter (Pall Corporation, New York, Supor filters to allow for analysis of diazotroph community NY, USA). The filters were placed in cryo tubes and composition. The filters were placed in sterile 1.5 ml cryo- immediately frozen in liquid N. vials containing 0.1 g autoclaved glass beads (BioSpec, Bartlesville, OK, USA), immediately flash frozen in liquid Sampling of individual particles and imaging N and stored at −80 °C until DNA extraction.

Individual particles (220 in total; ~20–500 μmindiameter) DNA extraction and particle lysis were visualized using a Nikon AZ100 microscope and a Nikon NI-150 illuminator and images were acquired using a DNA was extracted from seawater and bulk particle sam- Nikon DS-Fi1 camera. Individual particles were handpicked ples using the Qiagen Plant Minikit and QIAcube (Qiagen, using a micropipette and placed in PCR tubes. The pipette was Valencia, CA, USA). The filters were thawed at room rinsed with 2 µl sterile filtered 150 m seawater and the rinse temperature and 400 µl AP1 buffer was added to the sample was placed together with the particle in the tube. The tubes tubes. The samples were freeze-thawed three times using were frozen at −80 °C. In parallel, individual particles were liquid N and a 65 °C water bath and shaken at full speed for placed on glass slides, fixed in 2% paraformaldehyde and 2 min in a FastPrep-24 bead beater (MP Biomedicals, mounted using ProLong Diamond with DAPI (Molecular Irvine, CA, USA). The samples were Proteinase K treated probes, Eugene, OR, USA). Images of DAPI stained particles (45 µl of 20 mg ml−1 Proteinase K) for 1 h at 55 °C with and chlorophyll a containing cells were obtained using a Zeiss moderate shaking and RNase A treated for 10 min at 65 °C. Axioplan 2 and a Zeiss AxioCam HRc camera and the soft- The filters were removed from the sample tubes using sterile ware Zeiss AxioVision Rel.4.8. The filter cube for the chlor- needles and the further extraction steps were carried out ophyll a detection was a Zeiss high efficiency filter set 09, according to the manufacturer’s protocol. Samples of indi- with excitation BP at 450–490 nm and emission at LP515 nm. vidual particles were prepared for PCR amplification by To allow for a comparison between the particle samples subjecting them to three freeze-thaw cycles using liquid N and the surrounding seawater, seawater samples were col- and a 65 °C water bath, adding Lyse and Go PCR Reagent lected from 150 m (Fig. 1, S1 and S2). Seawater (2 l) was (10 µl; Thermo Scientific, Rockford, IL, USA) and placing filtered gently through a 10 μM polycarbonate filter (Ster- the tubes in a thermocycler with a program of 30 s at 65 °C, litech, Kent, WA, USA) or a 3 μM polycarbonate filter 30 s at 8 °C, 90 s at 65 °C, 180 s at 97 °C, 60 s at 8 °C, 180 s H. Farnelid et al. at 65 °C, 60 s at 97 °C, 60 s at 65 °C and an 80 °C hold genomics workbench (v7; CLC Bio, Qiagen, Boston, MA, while the PCR amplification mixtures were set up. USA) and sequences that were less than 350 bp in length and lacked either primer were discarded. Chimeric 16S rRNA and nifH gene amplicon sequencing and sequences were identified using the USEARCH [39] algo- sequence analysis rithm as compared with the GreenGenes 13.8 database [40] and filtered from the dataset. De novo operational taxo- A targeted amplicon sequencing (TAS) approach as nomic unit (OTU) clusters (97% sequence similarity) were described by Green et al. [35] was used to amplify the V3- generated, taxonomy of the OTU centroid sequence was V4 variable region of the 16S rRNA gene with the assigned using UCLUST in the QIIME v1.8 pipeline [41] CS1_341F (5′-ACACTGACGACATGGTTCTACACCTA and singletons were removed. Representative sequences for CGGGNGGCWGCAG-3′) and CS2_806R (5′-TACGG- each OTU were generated and closest relatives were iden- TAGCAGAGACTTGGTCTGACTACHVGGGTATCTA- tified using Blastn. The libraries were rarefied to 15 ATCC-3′) primers. The PCR was performed in 25 µl 000 sequence depth and samples with <15,000 sequences reaction volumes with 1× PCR buffer, 2.5 mM MgCl2, 200 were removed. The number of observed OTUs, Shannon H µM dNTPs, 0.25 µM of each primer and 0.3 µl Invitrogen and Chao1 diversity indices were calculated using Explicet Platinum Taq (Invitrogen, Carlsbad, CA, USA). The PCR [42]. A Bray–Curtis similarity matrix was calculated using conditions were 5 min denaturation at 95 °C, followed by 25 the software package Primer6 (Primer-E). To identify OTUs cycles of 40 s denaturation at 95 °C, annealing for 40 s at whose abundance in seawater differed significantly from 53 °C and elongation for 60 s at 72 °C, and ending with a that in bulk particle samples based on hierarchal taxonomy, final elongation for 7 min at 72 °C. For amplification of the a two-sided Welch’s t-test with confidence intervals was nifH gene, a nested PCR protocol was performed using the used in the STAMP v2.01 software [43]. Statistical con- nifH3 and nifH4 primers in the first PCR followed by a fidence intervals were calculated for a 95% nominal cov- second amplification with nifH1 and nifH2 primers [36] erage. PERMANOVA analyses were done using the vegan with CS1 and CS2 linkers [37]. For the PCR reaction package in R, and function adonis using 999 permutations. mixtures 1× PCR buffer, 4 mM MgCl2, 200 µM dNTPs, 0.5 The paired-end FASTQ files from the nifH amplicon µM of each primer and 0.3 µl Invitrogen Platinum Taq sequencing were merged using the software package PEAR (Invitrogen) were used. The PCR conditions were 3 min [38]. The libraries were quality filtered (phred20) using the denaturation at 95 °C, followed by 25 cycles in the first QIIME pipeline [41]. A de novo chimera check and OTU amplification and 30 cycles in the second PCR of 30 s clustering (97% similarity) was done using USEARCH61 denaturation at 95 °C, annealing for 30 s at 55 °C for the [39], representative sequences were generated and single- first amplification and 57 °C for the second PCR and tons were removed from the dataset. The sequences were elongation for 45 s at 72 °C, and ending with a final elon- translated and aligned using MEGA6.06 [44]. OTUs with gation for 7 min at 72 °C. representative sequences containing stop codons or frame- For the seawater and bulk particle samples, 1 µl of the shifts were removed from the dataset, corresponding to <9% DNA extract was used as a template and PCR reactions of the total number of sequence reads. To identify UCYN-A were run in triplicates. For the individual particles, 2.5 µl of sublineages, all UCYN-A reads were extracted from the the lysed solution was used as a template. Negative dataset by clustering all reads (post-chimera check) at 95% extraction controls (blank filter) and negative PCR controls nucleotide (nt) identity after adding representative UCYN- (only water instead of template) were included and did not A sequences. Oligotypes were determined using the oligo- result in any visible amplification when 5 µl amplification typing pipeline created by Eren et al. [45] using the same product was loaded onto a 1.4% agarose gel. The amplicons entropy positions determined in Turk-Kubo et al. [46]. were submitted for sequencing to the DNA Services Representative nifH sequences were annotated based on (DNAS) Facility at the University of Illinois at Chicago. At nifH cluster classification [47] according to the recently the DNAS Facility an additional PCR amplification was described CART model [48] complemented by a phyloge- performed to incorporate barcodes and sequencing adaptors netic tree analysis to identify groups of cyanobacterial to the final amplicons. The sequencing was performed on an diazotrophs. The sequences have been submitted to NCBI Illumina MiSeq sequencer using standard V3 chemistry Sequence Read Archive (SRA) with accession number with paired-end, 300 bp long reads and demultiplexing of SRP111426. reads was performed on instrument. For the 16S rRNA gene libraries, the forward and reverse Abundance of diazotrophs FASTQ files were merged using PEAR [38]. To obtain high-quality data, the sequences were trimmed using a To enumerate diazotroph groups in the vertical profile, quality threshold of p = 0.01 in the software package CLC quantitative polymerase chain reaction (qPCR) targeting Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre

UCYN-A1, UCYN-B, Trichodesmium spp., and three Results diatom–diazotroph associations (Het-1, Het-2, and Het-3) was used and conducted as previously described [49]. Water column features in the sampling area

The area of sampling was characterized by an anticyclonic eddy feature coinciding with elevated chlorophyll a and productivity (described in ref. [50]; Fig. 1). The mixed layer depth averaged around 30 m and the deep chlorophyll a maximum varied between 120 and 130 m. The surface photoautotroph population was dominated by Pro- chlorococcus at 1.6 × 108 cells l−1. Synechococcus and photosynthetic picoeukaryotes averaged 1.1 × 106 and 0.4 × 106 cells l−1, respectively [50]. The abundances of key diazotroph groups were determined, indicating that Croco- sphaera sp. (UCYN-B) was present in unusually high abundances in the upper 50 m (average 1.9 × 106 nifH gene copies l−1; Fig. 2). UCYN-A1, Trichodesmium spp., and three diatom-diazotroph associations (Het-1, Het-2, and Het-3) were also detected throughout the vertical proﬁle (5–100 m) albeit at lower abundances (Fig. 2).

Microscopic analysis of particles

Sinking particles were visualized and picked individually under a dissection light microscope. Many of the particles appeared to be loosely associated material, but fecal pellets and aggregates with a more dense appearance were also observed (Fig. 3). Filaments of Trichodesmium were observed in material from the net traps D1, D4, D5, and D6. Epiﬂuorescence microscopy showed that the particles were heavily colonized by photosynthetic (chlorophyll a con- Fig. 2 Average nifH gene abundances for target groups measured using qPCR in a vertical proﬁle. Error bars indicate standard devia- taining) and non-photosynthetic bacteria of various sizes tions based on triplicate measurements (Fig. 3).

Fig. 3 Microscopy images showing examples of individual particles from particle samples collected from net traps at 150 m. a–d The particulate material ranged from partly decayed planktonic organisms to loosely associated aggregates and fecal pellets. e, f DAPI stained particles showing that the particles were associated with both chlorophyll a containing cells (red) and non-pigmented cells (blue) H. Farnelid et al.

Fig. 4 Principal component analysis (PCA) plot calculated using the sequences in each library and unclassiﬁed sequences were retained. 16S rRNA gene class level of the taxonomical hierarchy for each Each sample is represented by a circle, with a color indicating the net sample. The two ﬁrst principal components are plotted with the pro- trap number or the fraction of the seawater. Bulk particle samples for portion of variance explained by each component indicated in brack- each net trap are labelled with a (+) in the circle ets. The numbers of sequences were normalized to the total number of

16S rRNA gene composition on particles seawater samples (Fig. 5). At the fine-scale OTU level, 24 OTUs with >1% difference in mean proportion were iden- The bacterial composition of the bulk particle samples, tified (Table 1). Typically free-living genera such as Syne- individual particles and seawater samples from the depth of choccoccus, Prochlorococcus and Pelagibacter were deployment (150 m) was investigated. In total, >2 million significantly (p < 0.01; Welch’s t-test) enriched in the sea- high quality 16S rRNA gene sequences were analyzed water samples compared to the bulk particle samples while (Supplementary Table S2). The lowest diversity was seen OTUs within the order Alteromonadales (Alteromonas; for the individual particles where the number of OTUs OTU_4074 and Pseudoalteromonas; OTU_26436) were varied between 387 and 875 OTUs. The highest diversity of significantly (p < 0.01; Welch’s t-test) enriched on particles the bacterial community was observed in the seawater (Table 1). samples (up to 1706 OTUs; Supplementary Figure S1). The relative abundance of chloroplast sequences was Principal coordinate analysis showed that the bacterial significantly higher in the bulk particle samples compared to community composition of the particles on both the class the seawater samples (p < 0.001; Welch’s t-test) and the and OTU level was most similar to the >10 and >3 µm most frequently detected chloroplast sequences were Stra- seawater samples; Fig. 4 and Supplementary Figure S2). menopiles (2.0 ± 2.2% mean sequence proportion) and The bulk particle samples exhibited less variation in com- Haptophycea (0.5 ± 0.3% mean sequence proportion; data position compared to the individual particles and were not shown). Notably, the chloroplast sequences of Rhizo- significantly different from the 0.2–10 and 0.2–3 µm sea- solenia imbricate (OTU_16894), a diatom which has been water samples (PERMANOVA, p = 0.005; Fig. 4 and reported to contain the endosymbiotic N2-fixing bacterium Supplementary Figures S2 and S3). The individual particles Richelia intracellularis [51], were among the most highly were highly variable in composition but a few particles enriched OTUs in the bulk particle samples (Fig. 6 and collected from the same net trap had relatively high simi- Table 1). In addition, OTUs closely related (>97% identity) larity (>75% Bray–Curtis similarity; Supplementary to the 16S rRNA gene sequences of the diazotrophs Cro- Figure S3). cosphaera, Trichodesmium and UCYN-A were consistently To identify groups whose abundance differed sig- present and significantly enriched in the bulk particle nificantly between the small size fraction of the seawater samples compared to the seawater samples (Fig. 6). samples (0.2–10 and 0.2–3 µm) and the bulk particle samples, a group comparison between the mean proportions of Presence and composition of diazotrophs on sequences per sample was done both at class and OTU particles level. The particle sequence libraries were dominated by Gammaproteobacteria (Fig. 5), including Alteromonadales, To identify the diazotrophs present on particles the nifH Vibrionales and Oceanospirillales. Bacteroidetes was the gene was amplified from bulk particle and individual par- second most abundant phylum, and Flavobacteria were ticle samples. All of the bulk particle samples had strong significantly (p < 0.001; Welch’s t-test) enriched in the nifH gene amplicons, visualized through gel electrophor- particle samples compared to the small size fraction esis. For the individual particles, nifH genes amplified from Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre

Fig. 5 Bar plot showing the mean proportion and the difference in mean proportion (16S rRNA gene) between the respective groups of signiﬁcantly enriched (p < 0.05; Welch’s t-test) classes with a difference in mean proportions of >1%. Error bars indicate standard deviation and corrected p-values are indicated to the right. The small size fraction seawater samples (0.2–10 and 0.2–3 µm) are shown in purple, and bulk particle samples from each net trap (D1–D7) are shown in green

49 out of the 76 individual particles (Supplementary Bray–Curtis similarity) while large variations between Tables S1 and S2). For the particles sampled from net traps individual particles were observed (Fig. 7 and Supplemen- D3 and D6, all resulted in nifH amplicons while for particles tary Figure S5). The composition of individual particles was from net traps D1 and D2 only <30% of the investigated highly dependent on the net trap sample from which the particles amplified. particles were collected. For example, Crocosphaera In total, >4.1 million high quality nifH sequences in 511 (UCYN-B; Cluster 1B) was most prevalent in individual OTUs (97% amino acid similarity clustering) were analyzed particles collected from the D3 and D5 traps (Fig. 7). (Supplementary Table S2). Classification of the nifH OTUs Similarly, OTU_1527 (Cluster III) which had 96% nt using a tree-placement approach showed that 44.5% of identity to a sequence reported from Seto Inland Sea in sequences in bulk particle samples were nifH Cluster 1B Japan (Accession number LC063947; [52]) and 99% amino Cyanobacteria (defined in [47]) with representative OTUs acid identity to Desulfovibrio putealis (Accession number from endosymbionts of diatoms (Het-1, Het-2 and Het-3), WP_027192250) was the dominating OTU in particles Trichodesmium, UCYN-A, Crocosphaera (UCYN-B) and collected from net trap D6 (up to 87 %). The OTU_1 Cyanothece (UCYN-C) (Supplementary Table S3). The (Cluster 1 P), affiliating with the Betaproteobacteria proportion of nifH Cluster 1B was lower in the bulk particle Azoarcus sp. and Dechloromonas aromatica (<88% nt samples compared to the vertical profile where Tricho- identity), was most frequently detected in individual parti- desmium, UCYN-A and UCYN-B were the most prominent cles collected from net trap D3 (Fig. 7). Three gammapro- 1B phylotypes. The UCYN-A sequences in the bulk particle teobacterial OTUs were represented among the eight most samples consisted primarily of the UCYN-A sublineages abundant OTUs including OTU_1107, representing the UCYN-A1 and UCYN-A3. The proportion of UCYN-A3 widespread gammaproteobacterial group γ24774A11 (98% sequences was consistently higher compared to UCYN-A1 nt identity; [28, 53]). However, although the gammapro- which has previously been reported as the dominant lineage teobacterial OTUs represented a large proportion of in the NPSG water column ([46]; Supplementary Fig- sequences in the bulk particle samples, their presence in ure S4). The nifH Cluster IG, a nifH cluster comprised individual particles was highly variable (Fig. 7 and Sup- primarily of gammaproteobacterial phylotypes was repre- plementary Table S3). sented by 50.3% of the bulk particle sequences (Supple- mentary Table S3). Sequences affiliating with nifH Cluster III, composed of diverse sequences mostly from anaerobic Discussion bacteria and archaea, Cluster 1P, which includes Gamma- and Betaproteobacteria phylotypes, and Cluster 1J/1K, Large diameter net traps allow for collection of sinking which includes Alpha- and Betaproteobacteria were con- particles over comparatively short timescales (4–7h). sistently recovered at a higher relative abundances in the Over the time course spanning 48 h, when our seven net bulk particle samples compared to the vertical profile traps were deployed and collected, the community com- (Supplementary Table S3). position of the bulk particle samples were relatively The nifH composition of the bulk particle samples was similar to one another (Fig. 4 and Supplementary Fig- highly consistent during the 48 h sampling, (76.6–90.5% ure S2). This suggests that there is a temporal consistency H. Farnelid et al.

Table 1 Affiliation and closest OTU ID Class affiliation Closest relative Similarity Accession relatives in Genbank to the number enriched (>1% difference in mean proportions) 16S rRNA Enriched in bulk particle samples gene OTUs in bulk particle samples (D1–D7) compared to OTU_43683 Alphaproteobacteria Shimia sp. 100% KT720466.1 seawater samples (0.2–10 and OTU_20532 Alphaproteobacteria Tropicibacter multivorans 100% KP843700.1 – 0.2 3 µm) OTU_16894 Chloroplast Rhizosolenia 97% KJ958482.1 OTU_25878 Flavobacteriia Unclassified Bacteroidetes; Flavobacteriales OTU_4074 Gammaproteobacteria Alteromonas sp. 100% KX356439.1 OTU_26436 Gammaproteobacteria Pseudoalteromonas sp. 100% KX806641.1 OTU_22018 Gammaproteobacteria Unclassified Gammaproteobacteria; Oceanospirillales OTU_29878 Gammaproteobacteria Thalassolituus oleivorans 97% KF170318.1 OTU_18280 Gammaproteobacteria Photobacterium angustum 100% KC534344.1 OTU_44280 Opitutae Uncultured Verrucomicrobia; 99% HQ675288.1 Puniceicoccaceae OTU_8705 Saprospirae Unclassified Bacteroidetes; Saprospiraceae Enriched in free-living seawater samples OTU_19700 Acidimicrobiia Uncultured Actinobacteria 100% HQ675198.1 OTU_36708 Alphaproteobacteria Pelagibacteraceae 100% HQ675347.1 OTU_39648 Alphaproteobacteria Pelagibacter sp. 100% LN850157 OTU_30217 Alphaproteobacteria Pelagibacteraceae 99% JF488446.1 OTU_817 Alphaproteobacteria Uncultured Alphaproteobacteria 97% JF488548.1 OTU_18321 Alphaproteobacteria Pelagibacteraceae 99% HQ675689.1 OTU_26273 Deltaproteobacteria Uncultured Deltaproteobacteria; 99% U65908.1 Sva0853 OTU_6835 Deltaproteobacteria Uncultured Deltaproteobacteria; 100% HQ675271.1 Sva0853 OTU_31137 SAR202 Unclassified Chloroflexi OTU_15929 AB16 Uncultured SAR406 96% JF488613.1 OTU_23515 Synechococcophycideae Prochlorococcus 100% CP007753.1 OTU_23585 Synechococcophycideae Synechococcus sp. 100% JF306716.1 in the composition of the microbial community on sinking The 16S rRNA gene libraries indicated that the particles particles over diel timescales. Moreover, the bulk particle contained both large and small size classes of cyano- samples in this study were enriched in taxa known to be bacterial diazotrophs. This was further confirmed by found in association with particles and the composition amplification and sequencing of the nifH gene in a majority was similar to that recently reported for traps deployed at of the individual particles. Our results demonstrate that in various depths at Station ALOHA in the NPSG [3]. By addition to diazotrophs associated with larger cells such as contrast, the sequence libraries of individually picked the diatom-Richelia symbiosis [19] typically associated particles (~50–200 µm in diameter) showed large varia- with sinking particles, unicellular cyanobacterial diazo- tions between the samples indicating that the individual trophs were also contributing to direct POM export. Uni- particles are inhabited by specialized bacterial taxa and cellular cyanobacterial diazotrophs have previously been offer distinct ecological niches. Notably, the individual visualized in association with inert particles and in asso- particles collected over the size range and characteristics ciation with eukaryotic cells [55, 56]. However, the cellular that could be easily distinguished and picked using a strategies of unicellular cyanobacterial diazotrophs and micropipette were not representative of the bulk particle whether they contribute to N2 fixation rates in both the free- samples (Fig. 4 and Supplementary Figure S2). This living and particulate fractions has not yet been resolved. potentially underscores the importance of small particles Why cyanobacterial diazotrophs of diverse size classes are for carbon flux [54]. found in association with sinking particles—a process that Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre

Fig. 6 Group comparison of the sequence library proportions of OTUs S2; >10, 0.2–10, >3, and 0.2–3 µm) are shown in purple and the bulk related to the 16S rRNA genes of known diazotrophs (>97% identity) particle samples from each net trap (D1–D7) are shown in green. The in seawater and bulk particle samples. The bar plots show the pro- horizontal line in each graph shows the mean proportions of sequences portion of sequences for each sample. The seawater samples (S1 and for the respective sample groups and p-values are given for each graph ultimately removes them from the euphotic zone—and the unicellular diazotroph Crocosphaera was present at whether these diazotrophs are active on particles have yet to unusually high abundances which likely reﬂected their be determined. presence in collected bulk particle samples. Some strains of The presence of unicellular cyanobacterial diazotrophs Crocosphaera are known to produce extracellular poly- on particles has implications for the particular lifestyles of saccharides [57, 58] which can contribute to particle for- speciﬁc diazotroph species as some may be more prone to mation and thereby they may have a direct role in carbon aggregation or grazing than others. At the time of sampling export [59]. The widely distributed, uncultivated, symbiotic H. Farnelid et al.

of heterotrophic N2 fixation; consequently, their significance to N2 fixation has been debated. It has previously been hypothesized that nutrient-rich particles could support heterotrophic N2 fixation [29, 61]. To our knowledge, this is the first study showing that diverse non-cyanobacterial diazotrophs are present on sinking particles which suggests that particles may be loci for heterotrophic N2 fixation. Particles collected from net trap D6 contained a large proportion of Cluster III diazotrophs and may therefore have exhibited conditions that favored anaerobic or microaerophilic bacteria. A central anaerobic micro-zone which can support anaerobic processes has been demonstrated in mm-large particles and aggregates [62, 63] and anaerobic or microaerophilic bacteria were recently reported in association with the particle size fraction (>3 µm) in the South China Sea [64]which suggests that bacteria with low O2 preferences could thrive on some particles. Cultivation studies of heterotrophic N2 fixation indicate that specific requirements of low O2 concentrations mayberequiredforN2 fixation [65]. However, it was recently shown that the sulfate-reducing N2-fixing bacterium, Desul- fovibrio magneticus, can grow in aerobic conditions [66]. Whether or not the particles that were picked in this study could provide the right conditions for Cluster III or other non- cyanobacterial diazotrophs to grow or fixN2 is unknown. Yet at times, the high proportion of anaerobic diazotrophs present within individual particles such as those from net trap D6 suggested that they might have been significant. Sequencing of gene amplicons is associated with several biases. For example, it has been suggested that the nifH primers used in this study may favor amplification of the gammaproteobacterial γ24774A11 group [67], represented by OTU_1107 in this study, while the endosymbionts Fig. 7 Bar plot showing the proportion of nifH sequences of the nine Richelia and Calothrix appear to be systematically under- most abundant OTUs (representing ~60% of the total nifH sequences represented in sequence libraries [49]. Notably, in this study in the dataset) for each of the individual particles and bulk particle although represented in the 16 S rRNA gene libraries (up to samples from each net trap (D1–D7). The legend indicates the nifH cluster affiliation and the OTU ID for each of the OTUs 0.9%, Fig. 6), the mean proportion of endosymbiont nifH groups (Het-1, Het-2 and Het-3) was low compared to other N2-fixing cyanobacterium UCYN-A has distinct co- groups in the bulk particle samples (5.0 ± 1.8%). This occurring sublineages, and in the NPSG, the two com- highlights that the differences in relative abundances monly occurring sublineages are UCYN-A1 and UCYN-A3 between taxa need to be interpreted cautiously. In addition, [46, 60]. While the symbiotic partner of UCYN-A3 has not the sequencing of a DNA fragment does not provide yet been identified or visualized, UCYN-A1 lives in asso- information of whether the cell is alive or even intact, thus ciation with a small haptophyte related to Braarudosphaera from sequencing data alone it cannot be determined whether bigelowii and the size of the cell consortia is 1–3µmin the detected diazotrophs were active on particles or not. In diameter [60]. The enrichment of UCYN-A3 on particles future studies, measurements of absolute abundances and suggests that these organisms aggregate on particles more N2 fixation rates will be necessary to elucidate the sig- efficiently or have adapted to a particle attached lifestyle. nificance of diazotrophs on particles but this may be tech- Thus, the strategies of the two closely related and co- nically challenging because of the heterogeneous nature of occurring sublineages may be different. the particle pool [54]. A large proportion of the sequences in the nifH gene To our knowledge, this is the first study which char- libraries in this study were from heterotrophic (non-cyano- acterizes the bacterial composition and presence of diazo- bacterial) diazotrophs. In the oligotrophic ocean, carbon trophs on individual sinking particles collected in the ocean. substrates are sparse and are unlikely to support the demands The study shows that the heterogeneity in composition on Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre particles is significant and indicates that diazotrophs are 7. Deng W, Cruz B, Neuer S. Effects of nutrient limitation on cell directly involved in POC export in the NPSG. The consistent growth, TEP production and aggregate formation of marine – presence of various size classes of diazotrophs warrant further Synechococcus. Aquat Microb Ecol. 2016;78:39 49. fi fi 8. Lomas MW, Moran SB. Evidence for aggregation and export of investigations, including quanti cation of speci c species and cyanobacteria and nano-eukaryotes from the Sargasso Sea measurement of N2 fixation activity on particles. Further euphotic zone. Biogeosciences. 2011;8:203–16. studies on microscale habitats will provide clues about the 9. Close HG, Shah SR, Ingalls AE, Diefendorf AF, Brodie EL, processes occurring within sinking particles and the important Hansman RL, et al. Export of submicron particulate organic matter to mesopelagic depth in an oligotrophic gyre. Proc Natl consequences for POC export in the oligotrophic ocean. Acad Sci USA. 2013;110:12565–70. 10. Richardson T, Jackson G. Small phytoplankton and carbon export from the surface ocean. Science. 2007;315:838–40. Acknowledgements We thank the C-MORE Team and the captain 11. DeLong EF, Franks DG, Alldredge AL. Phylogenetic diversity of and crew of the R/V Ka’imikai-O-Kanaloa and HOE-Legacy 2B aggregate-attached vs. free-living marine bacterial assemblages. (KOK1507), as well as Chief Scientist Tara Clemente at the Uni- Limnol Oceanogr. 1993;38:924–34. versity of Hawaii for assistance with deploying and sampling particle 12. Kellogg CTE, Deming JW. Comparison of free-living, suspended net traps. We thank Benedetto Barone for construction of a previous particle, and aggregate-associated bacterial and archaeal commu- version of Fig. 1. We would also like to acknowledge Stefan Green at nities in the Laptev Sea. Aquat Microb Ecol. 2009;57:1–18. the DNAS Facility at the University of Illinois at Chicago for the 13. Rath J, Wu KY, Herndl GJ, DeLong EF. High phylogenetic Illumina MiSeq sequencing. This work was supported by the Swedish diversity in a marine snow-associated bacterial assemblage. Aquat Research Council VR 637-2013-7502 to HF. The study is a con- Microb Ecol. 1998;14:261–9. tribution of the Simons Collaboration on Ocean Processes and Ecol- 14. LeCleir GR, Debruyn JM, Maas EW, Boyd PW, Wilhelm SW. ogy (SCOPE). Temporal changes in particle-associated microbial communities after interception by nonlethal sediment traps. FEMS Microbiol Compliance with ethical standards Ecol. 2014;87:153–63. 15. Pelve EA, Fontanez KM, DeLong EF. Bacterial succession on sinking particles in the ocean’s interior. Front Microbiol. fl fl Con ict of interest The authors declare that they have no con ict of 2017;8:1–15. interest. 16. Buesseler KO, Lamborg CH, Boyd PW, Lam PJ, Trull TW, Bidigare RR, et al. Revisiting carbon flux through the Ocean’ s Twilight Zone. Science. 2007;316:567–70. Open Access This article is licensed under a Creative Commons 17. Böttjer D, Dore JE, Karl DM, Letelier RM, Mahaffey C, Wilson Attribution 4.0 International License, which permits use, sharing, ST, et al. Temporal variability of nitrogen fixation and particulate adaptation, distribution and reproduction in any medium or format, as nitrogen export at Station ALOHA. Limnol Oceanogr. long as you give appropriate credit to the original author(s) and the 2017;62:200–16. source, provide a link to the Creative Commons license, and indicate if 18. Berthelot H, Moutin T, Helguen SL, Leblanc K, Hélias S, Grosso changes were made. The images or other third party material in this fi ’ O, et al. Dinitrogen xation and dissolved organic nitrogen fueled article are included in the article s Creative Commons license, unless primary production and particulate export during the VAHINE indicated otherwise in a credit line to the material. If material is not ’ mesocosm experiment (New Caledonia lagoon). Biogeosciences. included in the article s Creative Commons license and your intended 2015;12:4099–12. use is not permitted by statutory regulation or exceeds the permitted 19. White AE, Foster RA, Benitez-Nelson CR, Masqué P, Verdeny E, use, you will need to obtain permission directly from the copyright Popp BN, et al. Nitrogen fixation in the Gulf of California and the holder. To view a copy of this license, visit http://creativecommons. Eastern Tropical North Pacific. Prog Oceanogr. 2013;109: org/licenses/by/4.0/. 1–17. 20. Capone DG, Zehr JP, Paerl HW, Bergman B, Carpenter EJ. Tri- References chodesmium, a globally significant marine cyanobacterium. Sci- ence. 1997;276:1221–9. 21. Carpenter EJ, Montoya JP, Burns J, Mulholland MR, Sub- 1. Simon M, Grossart HP, Schweitzer B, Ploug H. Microbial ecology fi of organic aggregates in aquatic ecosystems. Aquat Microb Ecol. ramaniam A, Capone DG. Extensive bloom of a N2- xing diatom/ – cyanobacterial association in the tropical Atlantic Ocean. Mar 2002;28:175 211. – 2. Falkowski PG, Barber RT, Smetacek V. Biogeochemical controls and Ecol Prog Ser. 1999;185:273 83. – 22. Alldredge AL, Gotschalk C. In situ settling behavior of marine feedbacks on ocean primary production. Science. 1998;281:200 6. – 3. Fontanez KM, Eppley JM, Samo TJ, Karl DM, DeLong EF. snow. Limnol Ocean. 1988;33:339 51. Microbial community structure and function on sinking particles 23. Bar-zeev E, Avishay I, Bidle KD, Berman-Frank I. Programmed fi cell death in the marine cyanobacterium Trichodesmium mediates in the North Paci c Subtropical Gyre. Front Microbiol. – 2015;6:469. carbon and nitrogen export. ISME J. 2013;7:2340 8. 4. Guidi L, Chaffron S, Bittner L, Eveillard D, Larhlimi A, Roux S, 24. Moisander PH, Beinart RA, Hewson I, White AE, Johnson KS, Carlson CA, et al. Unicellular cyanobacterial distributions broaden et al. Plankton networks driving carbon export in the oligotrophic fi – ocean. Nature. 2016;532:465–70. the oceanic N2 xation domain. Science. 2010;327:1512 4. 25. Thompson AW, Zehr JP. Cellular interactions: lessons from the 5. Thiele S, Fuchs BM, Amann R, Iversen H. Colonization in the fi – photic zone and subsequent changes during sinking determine nitrogen- xing cyanobacteria. J Phycol. 2013;49:1024 35. bacterial community composition in marine snow. Appl Env 26. Foster RA, Kuypers MMM, Vagner T, Paerl RW, Musat N, Zehr – JP. Nitrogen fixation and transfer in open ocean diatom – cya- Microbiol. 2015;81:1463 71. – 6. Michaels AF, Silver MW. Primary production, sinking fluxes and nobacterial symbioses. ISME J. 2011;5:1484 93. the microbial food web. Deep Res. 1988;35:473–90. 27. Farnelid H, Andersson AF, Bertilsson S, Al-Soud WA, Hansen LH, Sørensen S et al. Nitrogenase gene amplicons from global H. Farnelid et al.

marine surface waters are dominated by genes of non- 46. Turk-Kubo KA, Farnelid HM, Shilova IN, Henke B, Zehr JP. cyanobacteria. PLoS One. 2011;6:e19223. Distinct ecological niches of marine symbiotic N2-fixing cyano- 28. Moisander PH, Serros T, Paerl RW, Beinart RA, Zehr JP. bacterium Candidatus Atelocyanobacterium thalassa sublineages. Gammaproteobacterial diazotrophs and nifH gene expression in J Phycol. 2017;53:451–61. surface waters of the South Pacific Ocean. ISME J. 2014; 47. Zehr JP, Jenkins BD, Short SM, Steward GF. Nitrogenase gene 8:1962–73. diversity and microbial community structure: a cross-system 29. Bombar D, Paerl RW, Riemann L. Marine non-cyanobacterial comparison. Environ Microbiol. 2003;5:539–54. diazotrophs: moving beyond molecular detection. Trends Micro- 48. Frank IE, Turk-Kubo KA, Zehr JP. Rapid annotation of nifH gene biol. 2016;24:916–27. sequences using classification and regression trees facilitates 30. Turk-Kubo KA, Karamchandani M, Capone DG, Zehr JP. The environmental functional gene analysis. Environ Microbiol. paradox of marine heterotrophic nitrogen fixation: abundances of 2016;8:905–16. heterotrophic diazotrophs do not account for nitrogen fixation 49. Turk-Kubo KA, Frank IE, Hogan ME, Desnues A, Bonnet S, Zehr rates in the Eastern Tropical South Pacific. Environ Microbiol. JP. Diazotroph community succession during the VAHINE 2014;16:3095–114. mesocosm experiment (New Caledonia lagoon). Biogeosciences. 31. Ploug H. Small-scale oxygen fluxes and remineralization in 2015;12:7435–52. sinking aggregates. Limnol Oceanogr. 2001;46:1624–31. 50. Wilson ST, Aylward FO, Ribalet F, Barone B, Casey JR, Connell 32. Ploug H, Bergkvist J. Oxygen diffusion limitation and ammonium PE, et al. Coordinated regulation of growth, activity and tran- production within sinking diatom aggregates under hypoxic and scription in natural populations of the unicellular nitrogen-fixing anoxic conditions. Mar Chem. 2015;176:142–9. cyanobacterium Crocosphaera. Nat Microbiol. 2017;2:17118. 33. Peterson ML, Wakeham SG, Lee C, Askea MA, Miquel JC. Novel 51. Villareal TA. Evaluation of nitrogen fixation in the diatom genus techniques for collection of sinking particles in the ocean and Rhizosolenia Ehr. in the absence of its cyanobacterial symbiont determining their settling rates. Limnol Oceanogr Methods. Richelia intracellularis Schmidt. J Plankt Res. 1987;9:965–71. 2005;3:520–32. 52. Hashimoto R, Watai H, Miyahara K, Sako Y. Spatial and temporal 34. Collins JR, Edwards BR, Thamatrakoln K, Ossolinski JE, variability of unicellular diazotrophic cyanobacteria in the eastern DiTullio GR, Bidle KD, et al. The multiple fates of sinking par- Seto Inland Sea. Fishe. 2016;82:459. ticles in the North Atlantic Ocean. Glob Biogeochem Cycles. 53. Langlois R, Großkopf T, Mills M, Takeda S. Widespread dis- 2015;29:1471–92. tribution and expression of gamma A (UMB), an uncultured, 35. Green SJ, Venkatramanan R, Naqib A. Deconstructing the poly- diazotrophic, γ -proteobacterial nifH phylotype. PLoS One. merase chain reaction: understanding and correcting bias asso- 2015;10:e0128912. ciated with primer degeneracies and primer-template mismatches. 54. Durkin CA, Estapa ML, Buesseler KO. Observations of carbon PLoS One. 2015;10:e0128122. export by small sinking particles in the upper mesopelagic. Mar 36. Zehr JP, McReynolds LA. Use of degenerate oligonucleotides for Chem. 2015;175:72–81. amplification of the nifH gene from the marine cyanobacterium 55. Biegala IC, Raimbault P. High abundance of diazotrophic pico- Trichodesmium thiebautii. Appl Env Microbiol. 1989;55:2522–6. cyanobacteria (<3 um) in a Southwest Pacific coral lagoon. Aquat 37. Moonsamy PV, Williams T, Bonella P, Holcomb CL, Höglund Microb Ecol. 2008;51:45–53. BN, Hillman G, et al. High throughput HLA genotyping using 56. Bonnet S, Biegala IC, Dutrieux P, Slemons LO, Capone DG. 454 sequencing and the Fluidigm Access ArrayTM system for Nitrogen fixation in the western equatorial Pacific: rates, diazo- simplified amplicon library preparation. Tissue Antigens. trophic cyanobacterial size class distribution, and biogeochemical 2013;81:141–9. significance. Glob Biogeochem Cycles. 2009;23:GB3012. 38. Zhang J, Kobert K, Flouri T, Stamatakis A. PEAR: a fast and 57. Sohm JA, Edwards BR, Wilson BG, Webb EA. Constitutive accurate Illumina Paired-End reAd mergeR. Bioinformatics. extracellular polysaccharide (EPS) production by specific isolates 2014;30:614–20. of Crocosphaera watsonii. Front Microbiol. 2011;2:229. 39. Edgar RC, Haas BJ, Clemente JC, Quince C, Knight R. UCHIME 58. Webb EA, Ehrenreich IM, Brown SL, Valois FW, Waterbury JB. improves sensitivity and speed of chimera detection. Bioinfor- Phenotypic and genotypic characterization of multiple strains of matics. 2011;27:2194–200. the diazotrophic cyanobacterium, Crocosphaera watsonii, isolated 40. McDonald D, Price MN, Goodrich J, Nawrocki EP, DeSantis TZ, from the open ocean. Environ Microbiol. 2009;11:338–48. Probst A, et al. An improved Greengenes taxonomy with explicit 59. Passow U. Transparent exopolymer particles (TEP) in aquatic ranks for ecological and evolutionary analyses of bacteria and environments. Prog Oceanogr. 2002;55:287–333. archaea. ISME J. 2012;6:610–8. 60. Farnelid H, Turk-Kubo K, Muñoz-Marín M, Zehr J. New insights 41. Caporaso JG, Kuczynski J, Stombaugh J, Bittinger K, Bushman into the ecology of the globally significant uncultured nitrogen- FD, Costello EK, et al. QIIME allows analysis of high- throughput fixing symbiont UCYN-A. Aquat Microb Ecol. 2016;77:125–38. community sequencing data. Nat Methods. 2010;7:335–6. 61. Riemann L, Farnelid H, Steward G. Nitrogenase genes in non- 42. Robertson CE, Harris JK, Wagner BD, Granger D, Browne K, cyanobacterial plankton: prevalence, diversity and regulation in Tatem B, et al. Explicet: graphical user interface software for marine waters. Aquat Microb Ecol. 2010;61:235–47. metadata-driven management, analysis and visualization of 62. Klawonn I, Bonaglia S, Brüchert V, Ploug H. Aerobic and microbiome data. Bioinformatics. 2013;29:3100–1. anaerobic nitrogen transformation processes in N2-fixing cyano- 43. Parks DH, Beiko RG. Identifying biologically relevant differences bacterial aggregates. ISME J. 2015;9:1456–66. between metagenomic communities. Bioinformatics. 63. Wilbanks EG, Salman-Carvalho V, Jaekel U, Humphrey PT, 2010;26:715–21. Eisen JA, Buckley DH, et al. The green berry consortia of the 44. Tamura K, Stecher G, Peterson D, Filipski A, Kumar S. MEGA6: Sippewissett salt marsh: Millimeter-sized aggregates of diazo- molecular evolutionary genetics analysis version 6.0. Mol Biol trophic unicellular cyanobacteria. Front Microbiol. 2017;8:1623. Evol. 2013;30:2725–9. 64. Zhang Y, Xiao W, Jiao N. Linking biochemical properties of 45. Eren AM, Sul WJ, Murphy LG, Grim SL, Morrison HG, Sogin particles to particle-attached and free-living bacterial community ML. Oligotyping: differentiating between closely related micro- structure along the particle density gradient from freshwater to bial taxa using 16S rRNA gene data. Methods Ecol Evol. open ocean. J Geophys Res G Biogeosciences. 2016;121: 2013;4:1111–9. 2261–74. Diverse diazotrophs are present on sinking particles in the North Pacific Subtropical Gyre

65. Farnelid H, Harder J, Bentzon-Tilia M, Riemann L. Isolation of bacteria in oxygen concentration gradient medium. Environ heterotrophic diazotrophic bacteria from estuarine surface waters. Microbiol Rep. 2016;8:1003–15. Environ Microbiol. 2014;16:3072–82. 67. Turk KA, Rees AP, Zehr JP, Pereira N, Swift P, Shelley R, et al. 66. Lefèvre CT, Howse PA, Lefe CT, Schmidt ML, Sabaty M, Nitrogen ﬁxation and nitrogenase (nifH) expression in tropical Menguy N, et al. Growth of magnetotactic sulfate-reducing waters of the eastern North Atlantic. ISME J. 2011;5:1201–12.